UK paid AI adoption more than doubled to 5.1% of adults, study finds
Bank-record research finds growth came mainly from new paying users, while Claude gained spending share partly through expensive plans. The figures exclude free use and employer-paid access.
The study estimates that 2.8 million UK adults personally paid for an AI assistant in June 2026, while growth in consumer spending has come mainly from adding subscribers rather than raising prices. For AI providers, the findings point to a market that is expanding but still limited in reach, with markedly different customer profiles across products. The estimates exclude free use and employer-paid licenses, and app-store subscriptions are partly inferred rather than directly identified.
01
Researchers used de-identified records from 14 million UK bank and card accounts, then weighted them to represent UK adults.
02
Among direct subscription payments, ChatGPT’s spending share fell from 73.5% to 52.5% between December 2025 and July 2026, while Claude’s rose from 7.4% to 29.4%.
03
Claude’s subscriber mix skewed toward expensive plans: 6.2% paid at least £80 a month, compared with 0.3% of ChatGPT subscribers.
A billion-pound consumer market, but still only one paying adult in 20: a new UK study puts AI’s commercial growth beside its limited paid reach. Researchers estimate that 5.1% of adults personally paid for an AI assistant in June 2026, more than double the 2.3% recorded a year earlier.
The November working paper, Who Pays for AI?, comes from Nikhil Datta and Peter John Lambert at the University of Warwick and Patrick Schneider at Imperial College London. Its estimates put the June paying population at about 2.8 million people. July consumer spending was equivalent to £1.07 billion a year if that month’s pace continued—not a tally of spending over a completed year.
The expansion was mostly about attracting customers rather than extracting more money from existing ones. The researchers attribute nearly nine-tenths of spending growth between January 2024 and July 2026 to more people paying, rather than higher spending per subscriber.
What a bank statement can—and cannot—reveal
The project draws on de-identified records from 14 million UK bank and card accounts. To identify direct purchases, researchers match statement descriptions, exact prices and launch dates against a registry of more than 200 AI products. They then weight the accounts to represent UK adults.
Purchases through Apple and Google’s app stores are harder to identify because the bank record does not name the product. Those subscriptions are estimated from recurring charges at AI subscription prices. The June adoption estimate therefore combines observed direct payments with estimated app-store subscriptions; its published 95% interval runs from 3.8% to 5.56% of adults.
This is a measure of personal purchases, not all AI adoption. Employer-paid licenses, business-account payments and free use fall outside the study. Someone using an assistant without paying personally would not count toward this estimate of paying adults.
Claude’s spending gains include a premium-plan effect
Within directly identifiable subscription payments, ChatGPT remained the leader, but Claude gained rapidly between December 2025 and July 2026. The comparison excludes app-store subscriptions. It also removes accounts at card issuers whose privacy filters hide Claude’s name, then reweights the remaining accounts to represent UK adults.
Editorial illustration for UK paid AI adoption more than doubled to 5.1% of adults, study finds.
Share of direct AI subscription spending
ChatGPT73.5%52.5%ChatGPT: 73.5% to 52.5%Claude7.4%29.4%Claude: 7.4% to 29.4%
ChatGPT73.5% → 52.5%
ChatGPT’s share of direct consumer AI subscription spending fell from 73.5% to 52.5%.
Claude7.4% → 29.4%
Claude’s share of direct consumer AI subscription spending rose from 7.4% to 29.4%.
The spending shift is not a subscriber-count comparison. The paper finds that 6.2% of Claude subscribers pay for plans costing at least £80 a month, compared with 0.3% of ChatGPT subscribers. Researchers say Claude’s gains were driven partly by technical and professional workers buying its most expensive plans.
Personal subscriptions follow a sharp occupational divide
The June occupation figures count direct payments, rather than adding estimated app-store purchases. Researchers identify occupations through payments to professional bodies, regulators and trade unions, alongside employer records. They publish only groups with enough accounts, and exclude card issuers that hide Claude’s name.
IT professionals: 20.3% paid directly for an AI assistant.
University and college lecturers: 16.0%; finance and insurance professionals: 15.8%.
Electricians: 2.9%; nurses and midwives: 2.5%; social care workers: 1.7%.
Datta interprets the occupational pattern as evidence that people are buying AI partly for help at work. These are out-of-pocket payments, however, rather than employer purchases. In comments published by TechXplore, he said people whose jobs involve tasks AI can help with were more likely to pay themselves.
Age and income show substantial gaps too, although those measures track whether someone had ever paid by July, not whether they held an active subscription. Among adults aged 20–29, 17.9% had paid, against 3.0% of those aged 60 and over. The corresponding figures were 18.8% for monthly incomes above £4,000 and 10.0% below £1,500. Income bands were taken from records predating November 2022.
Geography adds another concentration: London had about 78 paying AI users per 1,000 adults in June, compared with 46 across the rest of the UK. The researchers also identify clusters in Cambridge, Oxford and Edinburgh.
Sources
whopaysfor.aiWho Pays for AI? Datta, Lambert and Schneider (2026)
techxplore.comWho is behind the UK's billion-pound AI boom?
whopaysfor.aiWho Pays for AI? Paid AI adoption, measured monthly
Reader comments
Newest comments first. Replies stay oldest first.